ChatGPT's Voice Feature Now Available on Desktop, Enabling Spoken Task Control

OpenAI rolled out voice capabilities to its ChatGPT desktop application on Thursday, building on the ChatGPT-Live voice models that debuted earlier in July. Users can now speak naturally to the app to command and control AI agents that handle complex tasks across their computers.

What's Included#

The desktop implementation of ChatGPT Voice integrates with both ChatGPT Work and Codex. Through the desktop version, the system can access websites and applications, and on macOS machines, Appshots allows ChatGPT to read screen content and alternative text descriptions.

The feature also supports interruption detection, meaning users can break in and adjust commands mid-conversation without starting over.

Desktop vs. Mobile#

While ChatGPT Voice initially arrived on mobile devices with improved conversation flow and interruption handling, that version lacked task execution abilities. The desktop upgrade fundamentally changes this—users can now issue intricate, multi-step instructions and have ChatGPT execute them without human intervention for each stage.

OpenAI demonstrated the capability through a developer scenario: a single voice command to create a new thread, open a pull request, and pinpoint a bug's cause. The iOS app can also remotely access Codex functionality through the desktop client.

Market Context#

Competitor Anthropic has similarly advanced its Claude voice mode, which now works across productivity tools including Gmail, Calendar, Slack, Notion, and Canva through its suite of language models.